Are you ready to manage AI agents?
Ten statements across the five things an agent needs from you. Answer for your organisation today. About four minutes, no login. You get a read on each of the five and the one to fix first.
Job For each agent we run, we could write its job down clearly enough that two people would describe it the same way.
Job We have defined what each agent must never do, not just what it should do.
Context We have given our agents the context that lives in people's heads (the exceptions, the 'we never do it that way' rules), not just the standard procedure.
Context When an agent hits an unusual case, it handles it the way an experienced employee would.
Autonomy Each agent's autonomy matches the stakes: it acts freely on cheap, reversible decisions and waits for a human on the costly, hard-to-undo ones.
Autonomy Our people can say, without checking, exactly which decisions each agent makes alone and which come to a human.
Tools Each agent has the access it needs to do real work, and no more than that.
Tools For any agent, we can say instantly what it can touch and where to look to see what it did.
Monitor Someone is named and responsible for checking each agent's output on a set rhythm, with the authority to change things when it drifts.
Monitor We would notice if an agent that was right six months ago had quietly started being wrong.